A Unified View of Case Based Reasoning and Fuzzy Modeling

The fuzzy systems modeling technique and the case based reasoning methodology are briefly described. It is then shown that these two approaches can be viewed in a unified way as essentially involving the same process, a matching step and a solution composition step. It is noted that in the typical case based reasoning application the solution composition step is more difficult because of the complexity of the associated action space. Two techniques are then suggested to help in the solution composition task in case based reasoning. The first, the weighted median, is shown to be useful in domains in which the action space consists of an ordered collection of alternatives. The second, a variation of reinforcement learning, is shown to be useful in domains in which the resulting actions involve a sequence of steps.